研讨会信息
🎤 Speaker: Prof. Xu Jia
Associate Professor,
School of Future Technology & School of Artificial Intelligence,
Dalian University of Technology
📰 Title: Controllable Visual Content Generation
⏰ Time: 1:30-2:20 PM, BJ Time
📆 Date: 5 Dec., 2025 (Fri.)
📍 Venue: Hall C, GZ Campus
Online Zoom link:
https://hkust-gz-edu-cn.zoom.us/j/96983408548?pwd=moZgiftuonqonX6aTbOjTPDbkKRCWe.1
Zoom ID: 969 8340 8548
Passcode: 823262
研讨会概要
Abstract
While generative models are transforming creative industries, precise control over visual content generation remains a significant challenge. In this talk I will detail our latest advancements in controllable generation from static to dynamic. We present techniques for customizing character identity, interactions, and motions for personalized content. In addition, we also introduce post-training methods for fine-grained control over image aesthetics and spatial layout. For video, we propose a method for jointly controlling object and camera trajectories to enhance temporal consistency. Finally, we explore a novel physics-informed framework to address the lack of physical realism in current video generation models, aiming to produce more plausible and trustworthy visual content.
分享者简介
Prof. Xu Jia
Associate Professor,
School of Future Technology & School of Artificial Intelligence,
Dalian University of Technology
Xu Jia is an Associate Professor at the School of Future Technology and the School of Artificial Intelligence, Dalian University of Technology. He has been selected into the "Yangtze River Scholar" Youth Program and is recognized as a Dalian High-level Talent. Before that, he worked as a Senior Researcher in Noah' Ark Lab in Huawei and SenseTime. He received his Ph.D. degree from PSI/VISICS group of KU Leuven under the supervision of Prof. Tinne Tuytelaars and Prof. Luc Van Gool. His research focuses on computer vision and generative AI. He has published over 50 full-length papers in top-tier (CCF-A) journals and conferences, with over 11,000 citations on Google Scholar. He has applied for or been granted over 20 domestic and international invention patents. His research has received positive recognition from well-known scholars, including a Nobel laureate, and he is listed among the world's top 2% scientists. His achievements have been recognized with numerous academic awards, including the CCF Natural Science Award (Second Prize, ranked first), Huawei's "Challenge Demands" Spark Award, and the winner of the CVPR Shape Recovery Challenge. He frequently serves as an Area Chair or Senior Program Committee member for top tier AI and computer vision conferences such as ICLR, ICML, AAAI, and ACM MM.
扫描加关注
获取更多AI学域消息
SCAN & FOLLOW US!